When providing an application to a population of users, it may be useful to allow users to consent to reporting anonymous usage statistics of the application so that these usage statistics may be used to improve the application. In this regard, an experiment to test different variations of the application can be conducted (e.g., by assigning different variations to different users). The anonymous usage statistics of the users who consented can be compared to one another in order to determine which variation provided better performance and therefore can be used to improve the application. However, the more experiments that are conducted and/or the fewer users there are may mean that a particular user can be inadvertently assigned a unique set of variations (e.g., different from other users' assigned variations), thereby potentially making this particular user more uniquely identifiable.